Self Forcing trains autoregressive video diffusion models by simulating the inference process during training, performing autoregressive rollout with KV caching. It resolves the train-test distribution mismatch and enables real-time, streaming video generation on a single RTX 4090 while matching the quality of state-of-the-art diffusion models.
Requirements
We tested this repo on the following setup:
Nvidia GPU with at least 24 GB memory (RTX 4090, A100, and H100 are tested).
Linux operating system.
64 GB RAM.
Other hardware setup could also work but hasn't been tested.
Installation
Create a conda environment and install dependencies:
Our model works better with long, detailed prompts since it's trained with such prompts. We will integrate prompt extension into the codebase (similar to Wan2.1) in the future. For now, it is recommended to use third-party LLMs (such as GPT-4o) to extend your prompt before providing to the model.
You may want to adjust FPS so it plays smoothly on your device.
The speed can be improved by enabling torch.compile, TAEHV-VAE, or using FP8 Linear layers, although the latter two options may sacrifice quality. It is recommended to use torch.compile if possible and enable TAEHV-VAE if further speedup is needed.
CLI Inference
Example inference script using the chunk-wise autoregressive checkpoint trained with DMD:
Note: Our training algorithm (except for the GAN version) is data-free (no video data is needed). For now, we directly provide the ODE initialization checkpoint and will add more instructions on how to perform ODE initialization in the future (which is identical to the process described in the CausVid repo).
Our training run uses 600 iterations and completes in under 2 hours using 64 H100 GPUs. By implementing gradient accumulation, it should be possible to reproduce the results in less than 16 hours using 8 H100 GPUs.
Acknowledgements
This codebase is built on top of the open-source implementation of CausVid by Tianwei Yin and the Wan2.1 repo.
Citation
If you find this codebase useful for your research, please kindly cite our paper:
@article{huang2025selfforcing,
title={Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion},
author={Huang, Xun and Li, Zhengqi and He, Guande and Zhou, Mingyuan and Shechtman, Eli},
journal={arXiv preprint arXiv:2506.08009},
year={2025}
}
